Document papers recorded in a recording medium such as a recording paper have a plurality of types. The plural types of document papers are respectively read by a scanner to generate document images. Each of the document images undergoes a predetermined process according to the type of the document. If a document image the document type of which is unknown is input, the document processing device identifies the document type of the document image and performs a predetermined process for the document image according to the result of the identification.
A method of automatically identifying document types collectively stored features of plural document types, for example, the sizes of recording media, the number of characters, and the like in a database in advance. If a document image the document type of which is unknown is input, the method compares features of the input document image with the features of document types stored in the database, and searches pertaining document types. A technique regarding a method of identifying documents is disclosed in, for example, JP-A-2001-318941.